How to use BaseXClassifier properly ?

[LucasDresl]

Description

I am trying to implement BaseXClassifier for a classification setting and when i used BaseXClassifier.fit() on a binary dependant variable (y) throws ValueError: Unknown label type: 'continuous'

To Reproduce

from causalml.inference.meta import BaseXClassifier , BaseXRegressor , BaseXLearner
from causalml.dataset.classification import make_uplift_classification 

from sklearn.model_selection import train_test_split

from sklearn.linear_model import LogisticRegression

df , features = make_uplift_classification(n_samples=1000,
                               treatment_name=['control', 'treatment1', 'treatment2', 'treatment3'],
                               y_name='conversion',
                               n_classification_features=10,
                               n_classification_informative=5,
                               n_classification_redundant=0,
                               n_classification_repeated=0,
                               n_uplift_increase_dict={'treatment1': 2, 'treatment2': 2, 'treatment3': 2},
                               n_uplift_decrease_dict={'treatment1': 0, 'treatment2': 0, 'treatment3': 0},
                               delta_uplift_increase_dict={'treatment1': 0.02, 'treatment2': 0.05, 'treatment3': 0.1},
                               delta_uplift_decrease_dict={'treatment1': 0., 'treatment2': 0., 'treatment3': 0.},
                               n_uplift_increase_mix_informative_dict={'treatment1': 1, 'treatment2': 1, 'treatment3': 1},
                               n_uplift_decrease_mix_informative_dict={'treatment1': 0, 'treatment2': 0, 'treatment3': 0},
                               positive_class_proportion=0.65,
                               random_seed=20190101)

df_train , df_test = train_test_split(df , test_size = 0.2 , random_state = 44)

learner = BaseXClassifier(learner=LogisticRegression(),
                          control_outcome_learner=LogisticRegression(),
                          treatment_outcome_learner=LogisticRegression(),
                          control_effect_learner=LogisticRegression(),
                          treatment_effect_learner=LogisticRegression(),
                          ate_alpha=.05,
                          control_name='control')

learner.fit(df_train[features].values ,
                 treatment = df_train['treatment_group_key'],
                 y = df_train['conversion'].values) 

Environment:

• Mac
• Python 3.6.9
• pandas==1.0.5, sklearn==0.22.2

Expected behaviour
I was expecting to classifier fit because my y is binary (0 or 1)

[yungmsh]

Hey @LucasDresl thanks for the question- you need to supply a regressor model (e.g. LinearRegression) to the control_effect_learner and treatment_effect_learner arguments. FYI we're updating the API for next release whereby you just have to provide 2 models (instead of 4 currently), probably in a couple weeks.

[LucasDresl]

Hey @yungmsh ! thank you very much :) got it !

[zzlin1003]

Hi @yungmsh , can we use this BaseXClassifier to output the probability for each class, like the output of some sklearn classifier.predict_proba()?